Dynamic Deployment for User-Centric Multi-UAVs Communications with Fairness Consideration

In recent years, the use of unmanned aerial vehicles (UAVs) as temporary aerial base stations (ABSs) to provide services to mobile users has shown tremendous potential. However, traditional UAV-centric deployment strategy fails not only to fully leverage the flexibility of UAVs, but also to guarantee the fairness among multiple mobile users. Therefore, in this paper, a user-centric multi-UAVs communications network is constructed, where the adjacent mobile users are dynamically clustered, so that the position of each UAV ABS can be adaptively changed according to the time-varying user-centric clusters. To attain the fair throughput of the conceived network, a genetic algorithm assisted multi-agent deep reinforcement learning (GA-MADRL) method is proposed, which meticulously designs the positions of multiple UAVs and multiple access resources to enhance the user fairness of each constructed user-centric cluster. Simulation results show that, compared to the conventional UAV-centric deployment scheme, the proposed scheme significantly improves the fair throughput.